Bureaucrats or Politicians?∗

Alberto Alesina and Guido TabelliniHarvard University and Bocconi University

First Draft: May 2003This draft: July 2003

Abstract

Policies are typically chosen by politicians and bureaucrats.This paper investigates the criteria that should lead a society toallocate policy tasks to elected policymakers (politicians) or nonelected bureaucrats. Politicians tend to be preferable for tasksthat have the following features: they do not involve too muchspecific technical ability relative to effort; there is uncertaintyex ante about ex post preferences of the public and flexibility isvaluable; time inconsistency is not an issue; small but powerfulvested interests do not have large stakes in the policy outcome;effective decisions over policies require taking into account pol-icy complementarities and compensating the losers; the policiesimply redistributive conflicts among large groups of voters. Thereverse apply to the attribution of prerogatives to bureaucrats.

1 Introduction

Policies are chosen and implemented by both elected representatives(politicians) and non elected bureaucrats. The basic notion that politi-cians choose policies and bureaucrats execute them is overly simple. Theboundaries between decision and execution are a grey area and in manycases bureaucrats do much more than ”executing”. In most countriesnon elected central bankers conduct monetary policy, with much inde-pendence. Regulatory policies are normally the result of both political

∗For useful comments we thank Tim Besley, Andrei Shleifer, Charles Wyposzand participants at the CIAR meeting in Toronto, March 2003 and in a seminarin Geneva. This project was initiated while Alesina was visiting IGIER at BocconiUniversity; he is very grateful for the hospitality.

1

and bureaucratic intervention. Fiscal policy is, instead, by and large cho-sen by elected representatives (governments and legislatures). Foreignpolicy is decided by politicians. What criteria should lead a society toallocate decision power to politicians or bureaucrats on different policytasks? The goal of this paper is to try and answer this question, with anormative perspective. That is, we ask what is the optimal allocation oftasks at the constitutional table for society as a whole, and we only mar-ginally discuss the positive question of whether and how opportunisticpoliticians find it optimal to delegate to bureaucrats .Economists have emphasized one specific argument in favor of del-

egation of policy to a non elected bureaucrat: time inconsistency inmonetary policy, as in Kydland and Prescott (1977) and Barro and Gor-don (1983). Rogoff (1985) pointed out how an independent and inflationaverse central banker not subject to ex post democratic control wouldimprove social welfare. 1 But, as political scientists know well, timeinconsistency is clearly not the only criterion for choosing between bu-reaucrats and politicians. There is much more to it. For instance, evenfiscal policy is marred with a host of time inconsistency problems, butsocieties seem reluctant to allocate this policy prerogative to indepen-dent bureaucrats. 2 An ability to commit to a certain course of actionmay even be desirable in foreign policy, which however is always theprerogative of appointed politicians, at least in the more relevant phaseof choosing the general strategy. 3

Our starting point is the premise that politicians are motivated bya reelection goal, whereas bureaucrats are motivated by a ”career con-cern”. That is, politicians want to be re-elected; bureaucrats insteadwant to improve their professional prospects in the public or privatesector, and this motivates them to perform well whatever tasks theyreceive.4 In different circumstances and for different types of policies,these incentives play out differently and this leads to a variety of tradeoffs in the Constitutional choice of politicians or bureaucrats. That is,

1Walsh (1995) and Persson and Tabellini (1993) discussed ”contractual” arrange-ments between popular representatives and independent central bankers. For anempirical discussion of the benefits of independent central bankers see Grilli Mas-ciandaro and Tabellini (1991), Alesina and Summers (1993) for OECD countries andCukierman (1992) for a larger sample of countries.

2Note, however, that Blinder (1997) argues that some aspects of fiscal policycould be allocated to an independent agency operating like an independent CentralBank. Also the Business Council of Australia (1999) proposed that tax policy inAustralia be set by an independent agency within limits imposed by the legislature.

3See Putnam (1988) for a discussion of the role and benefits of commitments ininternational relations.

4For a discussion of how bureaucrats are motivated by prospect of career enhance-ment see the classic treatment in Wilson (1989) especially Chapter 9.

2

at the Constitutional table a society can choose which of these two in-centives (electoral and career) deliver superior results in different policyareas. We investigate both policies that have a ”public good” natureand redistributive policies. We also discuss the role of bribes and cam-paign contributions in alternative settings by stressing an agency prob-lem: politicians and bureaucrats need to be motivated to put effort orto abstain from grabbing rents for themselves.We analyze many different types of policies, trying to be reasonably

exhaustive. Politicians tend to be preferable for tasks that have the fol-lowing features: differences in performance are due to effort, more thanto individual talent or technical ability; the preferences of the public areunstable and uncertain, so that flexibility is valuable, a case that may beespecially relevant for complex policy environments; time inconsistencyis unlikely to be a relevant issue; the stakes for organized interest groupsare small, or the legal system is poorly designed so that corruption iswidespread; side payments to compensate the losers are desirable andrelevant, or bundling of different aspects of policy management and acomprehensive approach is important; and finally in policies that arepurely redistributive. The reverse applies to the attribution of preroga-tives to bureaucrats.These results seem consistent with some features of existing institu-

tional arrangements and may also serve as a normative benchmark forthe new Constitutions currently under formation such as the new Euro-pean Constitution and those of new democracies. A few examples mayclarify some of our points. Monetary policy involves fairly sophisticatedskills, has relatively few distributional consequences (compared say tofiscal policy) and social preferences on what is the appropriate goal ofmonetary policy do not vary much: at least ex-ante most people wouldagree that monetary policy ought to control inflation with some roomfor stabilization. Incidentally, these arguments provide a rationale forindependent central bankers even for those who do not believe that timeinconsistency of monetary policy is a major problem, such as Blinder(1999). On the contrary, foreign policy is an area where it is very diffi-cult to describe ex ante reasonably precise and fixed policy goals; in achanging world the preferences of the public may change substantially.Just think of how preferences for foreign policy changed in the US be-fore and after September 11, 2001. Finally, much of fiscal policy hasa redistributive nature and is in the hands of politicians. But otherredistributive policies that pit consumers at large against the special in-terests of monopolistic producers, such as regulation utilities, are oftencontrolled by non elected bureaucrats.Our paper has a normative focus, in that we ask how a Constitution

3

should allocate tasks between politicians and bureaucrats. In some re-spects, it provides a normative benchmark to the vast political scienceliterature asking the positive question of why legislative powers are del-egated in practice, what the effect of delegation is, the ”bureaucraticdrift” etc - see Epstein and O’ Halloran (1999). This literature, mostlyfocused on the US, is filled with interesting controversies. Some authors(Lowi (1969) amongst others) argue that delegation is deleterious, anabdication of the legislators’ responsibility and a way of favoring specialinterests (Stigler (1971)). Other authors (Mc Cubbins Noll and Wein-gast (1987 1989)) instead claim that the legislators can, at least up toa point control the bureaucratic agencies by means of procedural rules.But then, one may ask if control is imperfect, that is if the agencies canact following their own motivations, why delegate at all? One answer ismore ”optimistic” and relates to the need for division of labor, reductionof effort for the legislators etc.. Others are more cynical: Epstein and O’Halloran (1999) argue that the type of delegation chosen is the one thatmaximizes the benefits for elected politicians rather than social welfare.Fiorina (1977) point out the blame shifting role of delegation: politiciansdelegate to agencies in order to blame them when things go badly andclaim responsibility when success occurs.

Our paper is also related to several recent contributions that haveinvestigated the role of career concerns rather than explicit contracts.Dewatripont, Jewitt and Tirole (1999a,b) discuss the foundations of thisapproach and apply it to study the behavior of government agencies.They focus on some issues related to ours, namely the nature and ”fuzzi-ness” of the agencies mission, but they do not contrast bureaucratic andpolitical accountability. Dewatripont and Tirole (1999) study the roleof advocates that provide information and opinion to policymakers, anddiscuss how the career concerns of advocates may improve policymak-ing. Probably the paper more closely related to ours is Maskin andTirole (2001). Their goal is to investigate the attribution of responsi-bilities between accountable and non accountable agencies. The latterhave intrinsic motivations, while the former seek to please their princi-pals because of implicit rewards (career concerns). In our set up, in-stead, we neglect the role of intrinsic motivations: both bureaucrats andpoliticians need to be kept accountable with implicit incentives; but theimplicit incentive schemes can be of two kinds: those that define a politi-cian (striving for re-election), and those that define a bureaucrat (careerconcerns). Besley and Gathak (2003) also study intrinsically motivatedagents, and focus on how to combine intrinsic motivation with implicitrewards. Besley and Coate (2003) contrast appointed and elected regu-lators of public utilities; both policymakers’ types are intrinsically mo-

4

tivated, but direct election allows the voters to unbundle policy issues.Finally, we share with Aghion, Alesina and Trebbi (2002) a discussionof the role of electoral control versus delegation of power, although in arather different setting.The paper is organized as follows. Section 2 describes the simplest

case of our model and justifies its assumptions. Sections 3 and 4 discusscases of policies with a ”public good” nature and with no redistribu-tion. Section 5 discusses various issues arising when the policy decisionprocess is split between politicians and bureaucrats. Section 6 reviewsthe issue of delegation to solve time inconsistency problems. Sections7,8 and 9 deal with redistribution and with the role of organized interestgroups. The last section concludes.

2 The Model

Consider a society that has to decide whether to assign a certain policyto an elected officer or to a bureaucrat. With the generic term ”poli-cymaker” we indicate who chooses policy, so he or she can be either apolitician or a bureaucrat.

In the simplest possible case we consider a single policy, the resultof which is determined by the effort put in by the policymaker and byhis ability. Thus, the policy outcome y is defined as follows:

y = θ + a (1)

where a represents the effort of the policymaker and θ ∼ N(θ̄, σ2θ) is hisrandom ability. Ability and effort are additive.5 Citizens care about thepolicy outcome according to a well behaved, concave utility function,U(y). For the moment we consider linear preferences, U(y) = y, sincethe strict concavity of the utility function does not affect the natureof the results and simply makes the notation more cumbersome. Weintroduce strict concavity later when it matters.Effort is costly, and the strictly convex and increasing cost is labelled

C(a). The reward for the policymaker is labelled R(a) and it differsdepending on whether the policymaker is a politician or a bureaucrat.Both of them maximize their utility defined as:

R(a)− C(a) (2)

with Ca > 0, Caa > 0 and R(a) to be defined below (subscripts denotepartial derivatives).

5Alternatively they could be multiplicative leading to more complicated algebrabut similar results. See Dewatripont, Jewitt and Tirole (1999b) for a detailed dis-cussion of differences in the two formulations.

5

The timing is as follows. At the ”Constitutional Table” societychooses who has control rights over the policy (in the simplest case thereis only one, there will be multiple policies later). Then the policymakerchooses effort, a, before knowing his ability, θ. Finally nature chooses θ,outcomes are observed and the reward is paid. Irrespective of who hascontrol rights, only the outcome y is observable by the principals, notits composition between effort and ability. Hence the agent’s reward canonly be based on the policy outcome, y.Note that we assume that control over a policy con only be given

either to a bureaucrat or to a politician: we do not allow for joint controlover policies, or for some checks and balances between the two. Wereturn to this issue below in the discussion of extensions.

2.1 The bureaucrat’s rewardWe posit that bureaucrats are motivated by career concerns. That is,they are concerned with the perception of their ability θ in the eyes ofthose that may then promote them or offer them alternative job oppor-tunities in the private sector. Therefore the bureaucrat’s reward is (thesuffix B stands for Bureaucrat):

RB(a) = E(E(θ | y)) = E(y − ae) = E(θ + a− ae)

where ae is the public’s perception of a. Note that, if the principal is riskneutral, then the implicit reward offered by career concerns coincideswith the optimal contract for the principal, under the constraint thateffort and ability are not separately observable. In other words, if thebureaucrat were the agent of a principal who provides monetary reward,R(a) would be the equilibrium reward offered.

2.2 The politician’ s rewardThe politicians’s goal is to be reelected and he accomplishes this goalif y is above a threshold W. Therefore we have (the suffix P stands forPolitician):

RP (a) = Pr(y ≥W ) = 1− P (W − a) (3)

where P (W − a) = Pr(θ ≤W − a).We impose rationality of the voters,so that they expect that the alternative to reelecting the incumbent is toget another one with average talent, who in equilibrium will put the sameamount of effort as the current one.6 In fact every period is identicaland the politician’s effort choice is made before he observes his talent.It follows that:

W = θ̄ + ae (4)6Alternative assumptions about the determination of the threshold are of course

possible. We do not pursue here this extension.

6

We do not allow any career concerns for the politicians, other thanto be reelected. 7

3 Attribution of a single task

3.1 Perfect monitoringWe start with the simplest case, namely perfect monitoring of the policyoutcome y. We compute the first order condition with respect to effort,a, taking the expected level of effort ae as given, and then we impose theequilibrium requirement, that ae = a. For the bureaucrat we obtain:

1 = Ca(aB) (5)

where aB indicates the equilibrium effort of the bureaucrat.With a normal distribution for θ, equilibrium effort by the politician

aP , is defined implicitly by:

n(θ̄) = Ca(aP ) (6)

where n(θ̄) = 1/σθ√2π is the density of the normal distribution of θ

evaluated at its mean.It immediately follows that:

Proposition 1 aP ≶ aB depending on parameters’ values. But moreuncertainty about ability (a larger σθ) makes the bureaucrat more attrac-tive.

Intuitively, an increase in the variance of θ makes politicians put lesseffort. The reason is that voters’ behavior is less sensitive to good policyperformance, since more of the policy outcome is due to randomness.The marginal effect of effort on the perceived bureaucrat ability is insteadindependent of the variance of θ; this is because the compensation offeredto a bureaucrat is linear in performance, and thus independent of thedistribution of ability.One way of interpreting this result is that tasks for which talent is

especially important, that is tasks that are more inherently technical,should be allocated to bureaucrats rather than politicians a result thatseem reasonable.8 Similar results are obtained by Maskin and Tirole(2001) and by Epsetin and O’ Halloram (1999) in different models.

7Persson and Tabellini (2000) discuss the implications of this political model moreextensively. An alternative would be to allow a non-elected politician to also havecareer concerns, just as bureaucrats do. But as long as politicians retain an elec-toral incentive, this more general formulation would cloud the algebra without manyadditional insights.

8This result would be reinforced if the extent to which bureaucratic ability isrewarded were also allowed to vary. Tasks where technical abilities matter more arealso those for which rewards for ability are higher.

7

3.2 Imperfect monitoringWe now move to the case of imperfect monitoring, that is a situation inwhich performance is not perfectly observable. Thus, we add noise, ε,besides talent (θ) and effort (a) :

y = θ + ε+ a

with ε ∼ N(0, σ2ε), uncorrelated with θ and unobservable. Only perfor-mance y is observed and can be the basis of rewards.In this case the reward for bureaucrats can be rewritten as:

RB(a) = E(E(θ | y)) = θ̄ + βE(θ + ε+ a− ae − θ̄) (7)

where β = σ2θ/(σ2θ + σ2ε) < 1. Given our assumption of normality of the

distributions, we obtain a well known signal extraction result. Now theperception of talent is ”discounted” by a term β which reflects the signalto noise ratio. In equilibrium the choice of the bureaucrat is given by:

β = Ca(aB) (8)

Not surprisingly, the bureaucrat puts in less effort the lower is the signalto noise ratio.The politician reward is now given by the same expression as above,

except that now the distribution from which the probability Pr(y ≥W )can be computed has a larger variance, that reflects both the varianceof θ and of ε. It is immediate to derive the first order condition of thepolitician as follows:

n(θ̄, 0) = Ca(aP )

where n(θ̄, 0) = 1/(pσ2θ + σ2ε

√2π) is the density of the random variable

θ + ε, evaluated at the mean of both θ and ε.We are now ready to establish the following

Proposition 2 Imperfect monitoring (high σ2ε) reduces effort for bothtypes of policymakers. Higher σ2θ increases a

B but decreases aP .

Therefore, less monitoring does not favor one or the other type ofpolicymakers. However with imperfect monitoring a larger variance of θactually increases effort of the bureaucrat; therefore this result strength-ens what obtained in Proposition 1. This result is related to those ob-tained by Dewatripont, Jewitt and Tirole (1999b) who also point outthat performance less closely tied to talent or effort weakens the incen-tives of bureaucrats. But note that the same conclusions also apply toa politician. Hence, imperfect monitoring reduces the performance ofany policymaker, but it does not provide an argument for preferring apolitician to a bureaucrat at the constitutional stage.

8

4 Complex Tasks

We now add an element of complexity in tasks. In particular we focuson a situation (rather common) in which at the Constitutional Table thevoters are not sure how their preferences will evolve. We return to thecase of perfect monitoring and we assume that there are two possiblepolicies, that is two different directions in which effort can be devotedto, that is yi = θ + ai, with i = 1, 2.With multiple tasks, which will be our focus from now on, one

needs to specify a general cost function with multiple arguments, C =C(a1, a2). Instead of using the general formulation, we simplify to ei-ther an additive case (C = C(a1 + a2)) or to a separable case, (C =C(a1) + C(a2)) choosing the one that is simplest without generatingknife hedge or ”trivial” results. The more general specification of costsgenerates qualitatively similar results. We begin in this section by con-sidering additive costs, so that C = C(a1 + a2).At the Constitutional Table the (identical) voters are uncertain about

their ex post preferences over alternative policies, so that voters utilityis now given by the following strictly concave function:

U(λy1 + (1− λ)y2) (9)

with λ = 1 with probability q > 1/2, λ = 0 with probability (1 − q).Thus, society does not know ex ante what it will like ex post; but thereis no disagreement ex post amongst members of society. Disagreementsand redistribution will be analyzed below.The timing is now as follows. First, at the Constitutional Table the

voters choose whether to assign this policy to a bureaucrat or to a politi-cian, then nature chooses λ, that is social preferences are determined.Then the policymaker chooses [ai] , then nature chooses θ, and finallypolicy is determined and rewards paid. We assume that λ is observablebut not verifiable.9

Choosing a non-elected bureaucrat means that voters decide at theConstitutional Table to assign a task to the bureaucrat. Given that atthe Constitutional Table preferences are not yet known, one can onlyassign to the bureaucrat an unconditional task defined as follows:

y = δy1 + (1− δ)y2 (10)

where δ is a parameter specified by the Constitution. A crucial assump-tion is that the parameter δ cannot be contingent on the realization of

9Aghion Alesina and Trebbi (2002) also study of constitutional design in a casein which social preferences are not fully revealed ex ante. Their model and theiremphasis is however quite different.

9

the random variable λ : the mission for the bureaucrat cannot be con-tingent on the realization of ex post voters’ preferences. This elementof contract incompleteness is plausible: A bureaucrat is somebody whois not appointed through the political process, and therefore he will notfollow the ebb and flows of changing voters’ preferences. For examplethe independence of the central bank is tied to the fact that the centralbanker does not have to respond to the voters or even their represen-tatives for his policy choices, other than for how he fulfills the goalsassigned by the law to the central bank. But these goals can only be for-mulated in a simple and general way, like keep inflation under control;the central bank objectives cannot be changed with electoral results,almost by definition of what an independent central bank is. 10

Under these assumptions, the rewards of the bureaucrats are:

RB(a) = E(E(θ | y)) = E(θ + δa1 + (1− δ)a2 − δae1 − (1− δ)ae2) (11)

Given additive costs and q > 1/2, it is optimal to set δ = 1.11 The firstorder conditions for the bureaucrat imply:

aB1 = C−1a (1), aB2 = 0 (12)

That is the bureaucrat focuses all his effort on the ”main” activity of hismandate because that is more helpful in signaling his ability. Thus, thevoters’ utility in equilibrium is given by:

UB = qEU(θ + aB1 ) + (1− q)EU(θ) (13)

The key here is that by choosing a bureaucrat who is non responsive tothe ebb and flows of society’s preferences, citizens are ”stuck” with therisk that effort is misallocated and the bureaucrat pursues the wronggoals, those that ex-ante seem more likely to be relevant.This is what differentiates the politician from the bureaucrat. The

politician’s goals always depend on the realization of λ (i.e., on the pref-erences of the voters). Thus, knowing λ the politician will devote effortonly to the task preferred ex post by the voters according to the firstorder condition given above in (6). The following proposition follows.

Proposition 3 The politician always chooses the right task from thevoters’ perspective. This advantage of the politician is more im-portant the more risk averse are the voters.

10See Alesina and Gatti (1995) for an explicit discussion of insulation of the mis-sions assigned to the central bank form changes in the preferences of the electorate.11If costs were separable, then the optimal δ would be increasing with q, at a rate

that is decreasing with the curvature of U(.) for obvious reason having to do withrisk aversion;

10

This result is simple but important. Delegation to bureaucrats issafe when society’s preferences are well known and stable. But whenthey change, the ”rigidity” of a bureaucrat’s behavior makes the lattermuch less attractive. This helps us to understand why monetary policyis often delegated to an independent central bank, while foreign policy istypically under the control of politicians. Few would disagree with thestatement that the appropriate goal for monetary policy is to keep infla-tion under control with some room for stabilization policy; and this goalis unlikely to change over time. But preferences regarding foreign policyare unlikely to be stable and unchanged, and as a result an appropriatesimple bureaucratic goal cannot be stated once and for all. As a result,having a politician make decisions under direct democratic control andfollowing the ebb and flows of preferences in a changing world may besuperior to a fixed and unchangeable bureaucratic mission.Because in our formulation delegating policy to a bureaucrat implies

some rigidity, risk aversion plays a key role. More risk aversion impliesmore aversion to rigidity, a result which is also obtained in a differ-ent setting about constitutional design by Aghion, Alesina and Trebbi(2002).

5 Politicians and Bureaucrats

Thus far we have considered situations where the Constitution can onlyassign a task to either the politician or the bureaucrat. In this sec-tion we explore cases in which the interaction, that is a joint effort bypolticians and bureaucrats, may be superior to an either or arrangement.

5.1 Politicians decide, bureaucrats executeConsider again the case of complex tasks, as in the previous section. Anatural remedy to the "narrow-mindedness" of bureaucrats pursuing thewrong task is to let the politician decide the mission of the bureaucrat.Specifically, the constitution could prescribe that policy be delegatedto a bureaucrat, but the bureuacrat’s mission (the parameter δ in (10)above) be chosen by a politician. If the politician observes the contin-gency λ and if he is held accountable by the voters as described in theprevious section, he would always choose the socially optimal missionfor the bureuacrat. This division of tasks (the politician assigns the bu-reaucrat some goals and the latter chooses the instruments with whichto pursue them) is observed in a variety of real world arrangements. Anexample is the inflation targeting regime in the UK, where the govern-ment periodically assigns an inflation target to the Bank of England andthen it does not interfere with the central bank decisions of how to pur-sue that target. Of course, the precision and frequency with which the

11

goals of the bureaucrat are defined can vary from case to case, and deter-mine the extent to which an independent bureaucrat is really in chargeof policy decisions, (rather than taking orders from the politician). Inthe next sections we discuss more complex policy situations, in whichsociety may benefit from having really independent bureaucrats that donot take orders from politicians.Yet another possibility, more frequently observed in practice, is that

the constitution gives the politician the choice of whether or not to dele-gate at all, and if so how. Here one cannot be sure that the politician willdecide about the appointment of the bureaucrat in a way that maximizesocial welfare. The politician will make the choice that maximizes theprobability of reelection; whether or not this coincides with maximiza-tion of total welfare depends on the nature of the issues, the distributionof information, etc. In other words, this ”mixed”: arrangement givesrise to a tradeoff. The Constitution does not have to prescribe ex antea mixed (and ex post imperfect) task to the bureaucrat, but the de-cisions about whether or not to delegate is taken by selfish politiciansinterested in reelection rather than in social welfare. A recent literatureon whether or not and why Congress delegates addresses this tradeoff,and discusses when it is in the interested of elected politicians maximiz-ing chances of reelection to delegate to bureaucrats.(see Epstein and OHalloran (1999)).Whether or not maximization of the probability of victory of the

elected politician maximizes also social welfare depends on many dimen-sions; amongst other things, how the voters evaluate the politician whohas appointed a bureaucrat to perform a certain task. An interestingexample is the one of bureaucrats a "scapegoats". This may certainlyserve the opportunistic role of politicians, who taking advantage of theimperfect information of the public may indeed succeed, at least par-tially, in shifting the blame, as pointed out originally by Fiorina (1977).This may interfere with the public’s ability to appoint able and honestpoliticians.Note however that having scapegoats may actually be welfare en-

hancing because it may allow elected politicians to shift the blame for”unpopular” policies, for instance those with short run costs and longrun benefits. Since the politicians can shift the blame in equilibrium,one may have a higher dose of unpopular but welfare enhancing (in thelong run) policies, precisely because the electoral cost of these policesare reduced by blame shifting. Economists are familiar with how ”exces-sive” monetary tightening is often invoked by politicians to justify pooreconomic conditions even when monetary policy has nothing to do with

12

it.12 In Europe, often national politicians publicly blame bureuacrats inthe European Commission that tie their hands, but in private they some-times welcome these constraints and may even suggest how to formulatethe Commission’s recommendations. A similar role may be served byother international bureaucracies, such as the IMF when it prescribesso called ”unpopular” policies to macroeconomically unstable countries(see the discussion in Vreeland (2003)).

5.2 Politicians decide bureaucrats checkIn the previous subsection, we discussed how politicians may contributeto formulate the mission of the bureaucrats and thus to shape their in-centives. But the reverse is also often observed: sometimes an importantrole of bureaucrats is to provide information to the voters about whatpoliticians are relly doing. Since this information might change voters’behavior, it can also influence political choices. Auditors who scruti-nize government accounts are an obvious example, but not the only one.Often regulators make public statements about regulatory policy, andcentral banks do not refrain from making public comments about as-pects of fiscal policy.To illustrate this important role of bureaucrats, consider a version of

the model in section 4, but with the following changes. Task 1 providesshort term benefits to the voters (i.e. the voters observe y1 before theelections), while task 2 provides them long term benefits (here this meansthat y2 is observed by the voters only after the elections). As before,yi = θ + ai, i = 1, 2. The voters’ utility is now given by:

U(y1, y2) = y1 + (1 + λ)y2

where λ > 0 is a known parameter (this allows us to focus more simplyon the interesting case, but the model could be extended to a moregeneral set up). Voters reappoint the incumbent if

y1 ≥W = θ̄ + ae1 (14)

Note that voters’utility depends also on y2, yet their voting behaviorcannot be conditional on y2, since they observe it only after the elections.The politician is in charge of policy, and chooses how to allocate effort

between the two tasks, a1 and a2, and costs are additive, as in the previ-ous section: c = C(a1+a2). Here we assume that the politician has lexi-cografic preferences. First of all, he cares about his opportunistic benefits

12According to Alesina et al. (2001) for instance, this is precisely what has beenhappening in Europe since the creation of the European Central Bank, a very inde-pendent body, with an inflation target

13

net of costs, defined as usual, namely Max [Pr(y1 ≥W )− C(a1 + a2)] .But if he is indifferent between two actions that give him the same op-portunistic net benefits, then he also cares about voters’welfare.The role of the bureaucrat is to observe the task chosen by the politi-

cian and to inform the voters about it. The bureaucrat only observeswhether task 1 or 2 was selected by the politician, not the actual levelof effort. Here we simply assume that the bureaucrat’s incentive to tellthe truth are sufficiently strong. This assumption can be interpreted assaying that there is a small probability that the bureaucrat’s lies canbe detected ex-post, and this would destroy the bureaucrats’ long termcareer concerns. Of course, this assumption does not always make sense,and in some cases the bureaucrat could collude with the politician. Herewe only want to show how this framework can be generalized to studythe interaction between bureaucrats and politicians.To summarize, the timing of events is thus as follows: First, the

politician decides whether to allocate effort to the short term task, y1,or to the long term one, y2. Then the bureaucrat observes the taskchosen, and announces it publicly. Next, nature selects the politician’scompetence, θ, voters observe performance in task 1, y1, and electionsare held. Finally, the long term benefits y2 are enjoyed by the votersand the game ends.Since the voters know the bureaucrat’s announcement, their reser-

vation utility in the right-most side of (14) is conditional on what thebureaucrat announced, through the expectation term ae1. If the bureau-crat announces that the politician opted for the long term task, y2, thenae1 = 0. If instead the bureaucrat announces that the politician opted forthe short term task y1, then ae1 = a∗ > 0, where a∗ denotes equilibriumeffort. Thus, the bureaucrat’s announcement that the politician optedfor the short term task raises the reservation utility of the voters, andmakes re-election more difficult. Since the voters and the politician havethe same information, in equilibrium this leaves politician is exactly in-different between task 1 and 2, as far as his opportunistic benefits areconcerned. By the assumption that he has lexicographic preferences, inequilibrium he thus chooses the long term action y2 that is optimal forthe voters.This examples captures the idea that the bureaucrat’s longer term

career concerns can be usefully exploited to counter the politician’s ten-dency to deliver short term benefits to the voters. To achieve this, wedon’t really need to put the bureaucrat in control of policy decisions. Itis enough that he knows something about the true policy choices madeby the politician and acts as a watchdog. The voters’ behavior willthen endogenously adjust to induce an opportunistic politician to pur-

14

sue long term goals. Note that the voters are not deliberately punishingthe politician if he acts myopically. But they do so indirectly, whenthey exploit the bureaucrat announcement to better understand whatthe politician is really doing and use this information to reappoint theincumbent only if he is competent enough.

6 Time inconsistency

The benefits of flexibility associated with political delegation has a cost,when society’s preferences are time inconsistent. The rigidity of bu-reaucratic control, instead, offers protection against time inconsistency.Delegation to an independent agency to gain credibility is extensivelyused in monetary policy (as captured by Rogoff (1985)). Our modeloffers a different formalization of this point.Suppose, again, that there are two tasks, i = 1, 2, say fighting un-

employment (task 1) and enforcing tax collection (task 2). Citizens careabout both tasks, with simple linear preferences:

U(y1, y2) = y1 + y2 (15)

Effective tax enforcement depends on the policymaker’s effort and abil-ity, y2 = θ + a2. But equilibrium unemployment also depends on taxenforcement relative to the private sector expectations, ae2, because ofsome externalities. Specifically, suppose that the policy outcome in task1 (fighting unemployment) is given by:

y1 = θ + a1 − (a2 − ae2) (16)

Thus, low unemployment is brought about by ability and effort in choos-ing the right labor market policies (a1 + θ), but it is also facilitated byan unexpectedly low level of tax enforcement. For instance, the allo-cation of labor and capital between the underground and the formalsector depends on the expectation of tax enforcement (ae2). Tax enforce-ment amounts to discover and fine firms operating in the undergroundeconomy. The expectation of tighter enforcement helps the economy,because it induces firms to operate in the more efficient formal sector(this is where the externality would play a role). But once allocative de-cisions have been made by private individuals, tight enforcement maybecounterproductive (or less productive), because it subtracts resourcesfrom the private sector and it may force some firms operating in theunderground economy to close down or fire workers. Alternatively, asin the standard Barro and Gordon (1983) model of monetary policy,task 2 could be thought of as keeping low inflation, while task 1 couldbe stimulating economic growth through tax on structural policies; ef-fort in keeping inflation low (i.e. a high value of a2) reduces economic

15

growth, once private sector’s expectations of inflation have been freezedinto nominal wage contracts. Whatever the precise economic interpreta-tion, in this model the final outcome depends on the interaction betweenthe policymakers’ decisions and the private sector expectations, and thiscreates a time inconsistency.Suppose throughout that policy commitments are unavailable, mean-

ing that first private expectations are formed, and then effort in bothtasks, a1 and a2, are chosen. One can show (see the Appendix for thederivation) that politicians are much more likely to fall into the traps oftime inconsistency, compared to bureaucrats. The goals of a politicianare unavoidably linked to the ex-post welfare of voters, through reelec-tion motives. The bureaucrat instead can be given an explicit mission,possibly different from whatever is ex-post optimal for the voters. Thispossibility of strategic delegation enables society to overcome credibilityproblems. When time inconsistency is a prominent determinant of pol-icy decisions made under discretion, the case in favor of delegation toa bureaucrat is very compelling. This conclusion is essentially identicalto Rogoff’s point about strategic delegation in monetary policy. Butour framework shows more clearly another benefit of bureaucratic del-egation: it allows separation of tasks. One could assign one task to abureaucrat and one task to the politician. In the Appendix we showmore precisely the following:

Proposition 4 Under time inconsistency, the bureaucrat generally doesbetter than the politician, for two reasons: first, the mission of abureaucrat can be narrowly defined to avoid time inconsistent goals;second, even if this cannot be achieved because tasks cannot be splitamong separate agencies, the mission of a bureaucrat can be definedstrategically to influence private sector expectations, irrespective ofwhat is ex-post optimal for society.

A related issue has to do with the time dimension of the flow of costsand benefits of different policy tasks. As illustrated in the previoussection, politicians may have an incentive to devote a suboptimally highlevel of effort to tasks that give immediate benefits, visible before anelection. 13. What about the bureaucrat in the same situation? Tothe extent that their rewards have the same frequency of elections, theywould behave with the same incentive of politicians However, if theircareer rewards are more long term, they may have a lower incentive to tilttheir efforts toward short termist policies, for two reasons. One is that

13Models by Rogoff (1990), Rogoff and Sibert (1988) and Persson and Tabellini(1990 and 2000) are examples of this situation in the context of fiscal and monetarypolicy respectively.

16

often bureaucrats are appointed for longer than electoral cycles, oftenprecisely to avoid short termist policies.14 The second one is that even ifthe appointment of bureaucrats are short, if they distort policy choicesthe blame may reach them later on and harm them later. This gives thema stronger incentive to focus on the long term goal. A politician insteadmay be overwhelmingly interested in winning the next elections and beless worried about repercussions later on in his career, for two reasons.One is that winning the next election is the only thing that really mattersfor the politician, for instance because it is the last reelection opportunitythat he has. The second one is that even if his strategic manipulation ofeffort becomes known later on, it may not harm the politician that much.In future election the main issues at hand may be different and the votersmay forget past actions of the politician. On the contrary a bureaucrats’career may be much more sensitive to past mistakes. Thus, there is anargument for assigning to bureaucrats policy tasks that imply short termcosts and/or delayed benefits (or. as emphasized in the previous section,for using them as watchdogs of the politicians).15

7 Lobbying and bribing

We now turn to policies which imply conflicts amongst different mem-bers of society, broadly speaking redistributive policies with winners andlosers. In this section we consider the case of lobbies that can influencethe choice of policies with bribes or campaign contributions. Thus here”redistribution” is intended as favors toward powerful minorities. Theminority will seek to influence policy decisions to obtain favors. Boththe politician and the bureaucrat can be captured by the interest group,but with different mechanisms. This difference can give raise to a consti-tutional preference for one or the other type of policymaker, dependingon the circumstances. Note that a vast debate in political science, wellsummarized in Epstein and O Halloran (1999) informally discusses pre-cisely this point, namely whether delegation increases or decrease theability of special interests to receive favors at the expenses of society atlarge.There are two tasks, i = 1, 2, both affected linearly by effort and

ability, with no spillover effects across tasks: yi = θ + ai. The cost of

14Long terms of office for the Chairman of the Central Bank are considered anecessary tool to insure independence and a long term horizon in the conduct ofmonetary policy.15Besley and Coate (2003) find evidence that, in US states, elected regulators tend

to keep lower electricity prices compared to appointed regulators. If, as likely, lowerprices come at the expenses of lower investments, this finding is consistent with theprediction of short-termism by elected (as opposed to appointed) regulators.

17

effort is non-separable: C = C(a1 + a2). Task 1 benefits the voters atlarge, while task 2 only benefits a small but organized interest group.Voters influence policy only through elections. The organized interestgroup can influence policy either through bribes, b, or through campaigncontributions, f . Thus, the preferences of voters are just y1, while thoseof the interest group can be written as:

(1 + γ)y2 − b− f (17)

where γ is a parameter capturing the intensity of the group’s preferencesfor task 2.Bribes can be offered to both the politician and the bureaucrat, but

are illegal. Thus, if a policymaker accepts a bribe, with some exogenousprobability q he is caught and pays a fine Z (the interest group is notfined). Campaign contributions are legal and can only be offered to thepolitician. The effect of campaign contributions is to increase the incum-bent’s chances of winning the elections. We model this by saying thatthe voters’ reservation utility is a decreasing function of the campaigncontributions collected by the incumbent:

W = θ̄ + ae1 −H(f) (18)

where the functionH(.) captures the effect of the campaign contributionscollected by the politician. It is natural to assume that H(0) = 0, Hf >0, Hff < 0. Under these assumptions, we can write the policymaker’spreferences as:

R(y1, y2)− C(a1 + a2) + (1− φ)b− qZ (19)

where R(y1, y2) are the policymaker’s rewards (RB(y1, y2) = E(θ/y1)for the bureaucrat, RP (y1, y2) = Pr(y1 ≥ W ) for the politician), and1 > φ > 0 denotes transaction costs that reduce the value of the bribefor the recipient relative to the amount paid by the interest group. Thepolicymaker’s effort devoted to task 2 is observable by the interest group,so that bribes and campaign contributions can be contingent upon thepolicymaker effort: b = B(a2), f = F (a2). The timing of events is as fol-lows. First the Constitution allocates control rights over policies. Thenthe organized group commits to bribes and or campaign contributions,as a function of effort. Next, the policymaker allocates effort betweenthe two tasks. Nature then chooses a realization of θ. Finally, rewardsare paid.This is a common agency game, with two types of principals: the in-

terest group and the representative voter. The interest group has all thecommitment power and can either influence the agent directly (through

18

bribes), or indirectly (through campaign contributions). The distinc-tion between the politician and the bureaucrat is that the latter canonly be influenced by the interest group through bribes. We want toknow whether the voters are better off with the bureaucrat or with thepolitician, and what influences this comparison.

7.1 Bribing the bureaucratIf the constitution gave all control rights to the bureaucrat we wouldhave a standard common agency game, with a single active lobby. Ifbribes are positive, then the equilibrium must be jointly optimal for theorganized group and the politician. This immediately implies:

aB1 = 0, aB2 = C−1a (1 + γ) (20)

Moreover, restricting attention to truthful contribution (here brib-ing) schedules, the equilibrium bribing schedule has the following simpleform:16

B(a2) = B̄ +1 + γ

1− φa2 (21)

where the constant B̄ is chosen by the organized group so as to leave thebureaucrat indifferent between accepting or rejecting the bribe. Giventhe bureaucrat’s preferences, this implies:

B̄ = C(aB2 )− C(aB1 ) + aB1 − (1 + γ)aB2 + q̄Z (22)

where aB1 = C−1a (1) denotes the equilibrium policy if no bribe is accepted.Finally, the organized group must also prefer to pay the bribe rather

than be passive. This in turn puts an upper bound on the constant B̄that the organized interest group is willing to pay. Taking into account(22), an equilibrium with positive bribes exists only if the following con-dition is satisfied:

(1− 2φ)(1 + γ)

1− φaB2 −

£C(aB2 )− C(aB1 ) + aB1

¤ ≥ q̄Z (23)

If instead this condition is violated, then the equilibrium with the bu-reaucrat delivers the optimal policy for the voters. Equation (23) makesit clear that an equilibrium in which the bureaucrat is bribed is morelikely if the stakes for the organized group are high (γ is large), or if thelegal system works poorly (qZ and φ are small).

16See Grossman and Helpman (2001).

19

7.2 Lobbying the politicianNext, suppose that the politician is in charge of the policy decision. Acondition very similar to (23) above determines the existence of an equi-librium with bribes (the expression is not identical because the politi-cian’s reward occur through reappointment). In particular, it remainstrue that bribes would be zero if the legal system is strong, so that trans-action costs are high or the probability of being caught is high. But now,besides bribes, the organized interest group can also resort to campaigncontributions. He will choose to do so if campaign contributions aresufficiently effective in swaying the voters.Specifically, in an equilibrium with campaign contributions, the allo-

cation of effort must be jointly optimal for the politician and the orga-nized group. Thus, the equilibriummust solve the following optimizationproblem by choice of a1,a2 and f, subject to non-negativity constraintson the three choice variables, and taking voters’ expectations ae1 as given,as before.

Max©Pr(θ ≥ θ̄ + ae1 − a1 −H(f)) + (1 + γ)a2 − C(a1 + a2)− f

ª(24)

The properties of the solution to this problem depend on the slopeof the function H(f), i.e., on how effective campaign contributions arein swaying the voters. In the Appendix we consider two cases:First, if Hf(0) < 1/(1 + γ), then the equilibrium has zero lobbying

(f = 0) and the outcome is optimal for the voters (aP2 = 0). In this case,campaign contributions cannot be productive enough, and the organizedgroup will not seek to influence the politician: the group’s stakes aretoo low relative to how much he would have to pay into the electoralcampaign of the politician.The opposite extreme occurs if Hf(f

∗) > 1/(1 + γ), where f∗ de-notes equilibrium campaign contributions, to be defined below. In thiscase, campaign contributions are very effective at the margin. Effortis allocated so as to please only the organized group, as in (20) above.And equilibrium campaign contributions are defined implicitly by theoptimality condition:

n(θ̄ −H(f∗)) ·Hf(f∗) = 1 (25)

where n(x) is the normal density of θ evaluated at the point x. For thisto be an equilibrium, the organized group must benefit relative to theoption of not lobbying at all, and this also requires: (1 + γ)aP2 ≥ f∗.We summarize this discussion in the following:

Proposition 5 Political lobbying can be an equilibrium, even if bribesto the bureaucrat are not. This is more likely if campaign contri-

20

butions are effective in influencing the voters, but the legal systemis strong and effective in discouraging bribes.

Thus, politically appointed policymakers are more easily captured byorganized interests compared to bureaucrats, particularly in advanceddemocracies with a well functioning legal system. The reason is that,to influence a politician, the interest group needs to convince the votersthat the politician is doing a good job and deserves to be reelected. Thepolitician will then automatically respond with policy favors to the in-terest group, since this will help his chances of reelection. To influencea bureaucrat, instead, the organized group needs to engage in illegal orsemi-legal activities, and fight against possibly deeply entrenched pro-fessional goals and standards of a technical bureaucracy. Policies wherethe stakes for organized interests are very high, or where redistributiveconflicts concern small but powerful vested interests against the votersat large, may thus be more safely left in the hands of the bureaucrat.This conclusion might explain why, in many advanced countries, theregulation of public utilities is typically done by bureaucrats. In thesecases, the long run interests of consumers are easy to identify, and thestakes for the utilities’ supplier are very high, so that a politician maybe more easily captured than the regulator.17

8 Compensation of losers

One critical task for politicians is to form coalitions in favor of certainpolicies, compensating losers either with direct transfers or by bundlingseveral policies into one package. To illustrate this point, we need a con-flict of interest between voters (or groups of voters) and the possibilityof side payments and a possibility of bundling policies with complemen-tarities.Voters’ utility now depends on the policy outcome and the transfer

(positive or negative) received by the government. We have two voters( or homogeneous groups of voters of equal size) with concave utilityfunction defined over private consumption, U(ci), i = 1, 2 and where:

c1 = y1 + t, c2 = y2 − t, y2 ≥ t ≥ −y1 (26)

Therefore t is a direct lump sum transfer between voters and the gov-ernment budget is balanced; there are no tax distortions. Each groupbenefits from different tasks requiring specific and uncorrelated abilities,θi, i = 1, 2. Let the distribution of θi have the same densities n(.) and cu-mulative distributions N(.) (not necessarily normal). There are random

17See however the evidence by Besley and Coate (2003) quoted above.

21

negative spillovers between the two tasks, such that:

y1 = θ1 + a1 − λκa2, y2 = θ2 + a2 − (1− λ)κa1 (27)

The parameter 0 < κ < 1 denotes the strength of the negative spillovereffects. But who is hurt by the spillovers is ex ante uncertain. Thus, λis a random variable that can equal 1 or 0 with equal probabilities. As insection 4, we assume that λ is observable but it is not verifiable, so thatthe bureaucrat’s mission cannot be defined contingent on λ. Thus, thepolicymaker will maximize its usual payoffs, with additive cost functionsand with different rewards for the two types of policymakers:

R(a1, a2)− C(a1)− C(a2) (28)

Timing has the usual structure. First nature sets λ and this deter-mines which group is hurt by the spillover effect. Then the policymakerchooses ai and t, nature sets θi and rewards are paid.Consider the politician first. He maximizes reelection probabilities,

which means that he has to win the favor of a strict majority of voters.Here this means winning the votes of both groups (as it will be clear be-low, nothing of substance hinges on the fact that in this simple examplereelection requires pleasing all voters). Therefore:

RP (a1, a2) = Pr ob(U(c1) 1W1) ∗ Pr ob(U(c2) 1W2) (29)

where Wi is the reservation utility of group i.Suppose for concreteness that λ = 1. If the two reservation utilities

are equal, then the politician sets transfers t so that:

n(x1)

1−N(x1)=

n(x2)

1−N(x2)(30)

where x1 = U−1(W )−t−a1+κa2 and x2 = U−1(W )+t−a2. That is, thepolitician equalizes the ”hazard rates” of losing votes from either group.In this context, the hazard rate measures the elasticity of the probabilityof winning with respect to transfers. Thus, this optimality condition issimilar to the Ramsey rule of optimal taxation: transfers are allocatedbetween groups so as to equalize this elasticity across groups. If thehazard rate is monotonically increasing in x, and given the assumptionof the same distribution for θi, i = 1, 2, equation (30) implies c1 = c2.

18

That is, the politician implements full insurance, fully compensating thelosers from the negative externality (remember that compensations are

18A uniform distribution of θ satisfies the assumption of a monotonically increasinghazard rate, for instance.

22

costless, if they implied a transaction costs or a distortionary cost theequalization of utilities would no be complete).Exploiting (30), the optimality conditions for the allocation of effort

to the two tasks imply:

n(1−N2)=Ca(aP1 ) (31)

n(1−N1)(1− κ)=Ca(aP2 )

Thus, the politician will allocate effort ”correctly”, in the sense of de-voting more effort to the task that does not have negative spillovers:aP1 > aP2 if λ = 1. Comparing (31) with (6) in section 2, however, wesee that the politician is induced to put less effort also in the task withno negative externality (task 1), relative to the simple case of only onetask. The reason is that bundling of two tasks requiring different abili-ties weakens his incentives. His likelihood of reelection now depends onhis success in both tasks. Even if he puts a lot of effort in task 1, hecould still loose the election because he happens to be unable in task2. His awareness of this risk (captured by the term (1−N2) on the lefthand side of (31)), dilutes his incentives.19

Let’s now turn to the bureaucrat. By assumption, the goal he isassigned cannot be contingent on λ and has to be stated at the Consti-tutional Table. The natural goal is to maximize total output, (y1 + y2).If given this goal, the bureaucrat will allocate effort efficiently, takingthe negative externality into account:

1=Ca(aB1 ) (32)

1− κ=Ca(aB2 )

Nevertheless, compensating transfers will be set to zero.Comparing the politician and the bureaucrat, we thus have:

Proposition 6 The politician provides side payment to compensate losersbut has weaker incentives than the bureaucrat; the latter, however, doesnot compensate losers.

This result relies on the fact that bureaucrats cannot be given statecontingent missions, and if their goal is formulated in terms of aggre-gate efficiency, they will neglect the distributional consequences of theiractions. A politician instead can take advantage of relatively complexand evolving spillovers between issues and build majorities with com-plex side payments schemes. Compensating the losers makes it easier

19Persson and Tabellini (2000) and Seabright (1996) elaborate further on this pointcomparing centralized vs decentralized arrangements.

23

to pass legislation while at the same time providing insurance againstbad luck. Imagine a policy that favors a large majority, say a badlyneeded highway, but that creates losers, say the property owners. Underdemocratic choice, the losers might be able to block the project. But thepolitician can put together a package of compensation for the propertyowners, with large benefit for the majority. In a sense this is almostwhat describes the job of a politician. Instead, it is hard to imagine howa bureaucrat might do that. How can one write on paper what a bureau-crat is allowed to do or not do, to create bundling and compensation? Abureaucrat can be delegated the task of building the best possible high-way and he may potentially do a better job than the politician; but hemay not have the ability, interest or authority to provide compensationto the local owners. Note also that ”writing some checks” to compensategroups of losers does not require any particular technical competence,another reason why it may be difficult to generate the correct incentivesfor bureaucrats motivated by the career concerns that we have modeled,an observation which leads us directly into the next section.

9 Splitting the cake

We now consider a purely redistributive policy, ”cake splitting”. Con-sider three voters, the minimum number required to make the probleminteresting. The policy task delivers a ”cake” that can be divided be-tween the three voters, therefore:

y = θ + a = c1 + c2 + c3 (33)

The utility function of the voters is concave, U(ci), i = 1, 2, 3. The keydifference between a politician and a bureaucrat is, once again, that theformer needs a majority to win and the latter simply wants to signaltalent. The bureaucrat can either be given no redistributive tasks, inwhich case redistribution is entirely arbitrary; alternatively, behind a veilof ignorance he can be assigned the task of redistributing equally, thatis y/3 for all three voters. The politician needs two votes, and thereforehe will give y/2 to two voters and zero to the other one. For the sake ofexposition consider first the case of risk neutrality, i.e. U(ci) = ci.Consider the politician first. Since he only needs to please two voters

out of three, his reward is:

RP (a) = Pr ob(y ≥ 2W ) (34)

where W is the reservation utility of individual voters. With forwardlooking and rational voters, W equals the average expected utility theycan get if the opponent is elected. If the hypothetical redistribution

24

implemented by the opponent is unknown ex-ante, then W = 1/3(θ +ae). Going through the usual steps, of maximizing with respect toeffort for given expectations and then imposing rational expectations, inequilibrium the politician optimality condition implies:

n(2θ/3− aP ) = Ca(aP ) (35)

where n(x) denotes the normal density evaluated at the point x. Com-paring (35) with (6) in section 3, we see that once the politician is also incharge of redistribution, he can get away with less effort in equilibrium,compared to the case of no redistribution. The reason is that, as alreadystated, here he only needs to please two voters out of three. He can thusreduce effort, and still please two voters with the portion of the caketaken away from the minority.20

Next, consider the bureaucrat, and suppose that he is "fair", in thesense that he gives one third of y to each voter. Then his first orderconditions are still identical to those formulated in (5), section 3. Thismakes the bureaucrat more attractive for the voters for a larger rangeof parameter values, compared to the case of no redistribution. Intu-itively, if indeed the mission of the bureaucrat can be formulated so thathe abstains from redistributing and treats all voters equally, then thebureaucrat has a further advantage relative to the politician: he puts inmore effort, because his incentives are not weakened by the possibility ofredistribution. This advantage of the bureaucrat is reinforced if votersare risk averse, because then the politician would also expose the votersto the risk of being in the minority, compared to a "fair" bureaucrat.There is however an interesting time inconsistency problem. Sup-

pose that a bureaucrat has been given the task. Ex post, when thevoters learn which majority would be put together by the politician, thewinning majority of the voters would generally want to take the taskaway from the bureaucrat and to replace him with the politician.21 Ifthe redistributive stakes are strong enough, the constitution may nothave enough commitment power to resist these political pressures. Inthis case, delegation to a bureaucrat may not be feasible in equilibrium,even though it might be desirable ex-ante. This result underlies andan additional distortionary effect of redistributive policies, even without20This result is similar to that obtained in Ferejohn (1986) and Persson and

Tabellini (2000). But since here voters are forward looking, we rule out the Bertarandcompetition among voters that instead features in the backward looking voting equi-librium of Ferejohn (1986).21The precise conditions under which a majority of the voters would prefer to

replace the "fair" bureaucrat with the politician depend on the details of the timing,and in particular on whether the replacement occurs before or after effort has beenchosen by the bureaucrat.

25

tax distortions. Given an inability to commit to ex ante ”fair” policies,in equilibrium a society may end up with a smaller cake.Finally it is worth mentioning the case in which no redistributive

goals can be given to a bureaucrat and, therefore, he distributes thecake in a totally arbitrary fashion. It is easy to see that in the case ofrisk neutrality this does not affect the choice ex ante between bureau-crat and politician. But with risk aversion, uncertainty about how abureaucrat would allocate the cake increases the ex-ante desiderabilty ofthe politician. Redistribution under the politician is less risky, since twovoters out of three are always included in the winning majority.We can summarize the previous discussion into this

Proposition 7 The possibility of redistribution reduces the equilibriumeffort of the politician, but not that of the bureaucrat. Risk aversionmakes the bureaucrat less or more desirable ex-ante depending on howeasy it is to impose fair treatment of all voters in his task description.But even if the bureaucrat might be preferable ex ante, ex post a majorityof the voters would generally renege on this choice.

Summarizing, there are two reasons why politicians may be preferablefor splitting cakes, i.e. for purely redistributive policies. One is that atime inconsistency problem may make the choice of a bureaucrat notsustainable ex-post, even though ex ante optimal. The second reason isthat one can judge talent from the size of the cake, but it is more difficultto judge talent from how one cuts the cake. The constitution chosenbehind a veil of ignorance could give the bureaucrat precise directionsabout how the cake should be cut. In practice, however, it may be quitedifficult to describe ex ante a precise redistributive scheme. Leaving asidethe time inconsistency problem, risk averse voters may be unwilling todelegate redistribution to a bureaucrat interested only to maximize thecake and with unclear incentives about how to cut it.The result that legislators rather that bureaucratic agencies are in

charge of purely redistributive policies is similar to predictions of litera-ture well summarized by Epstein and O Halloran (1999). One argumentin this line of research is that legislators will never choose to delegatepolicies which are especially rewarding at the pools like redistributions.But this result has a ”positive” rather than a normative bend as in ourmodel.

10 Conclusions

Our analysis rests on two fundamental assumptions. The first one con-cerns the motivation of different types of policymakers, bureaucrats and

26

elected politicians. The former wants to signal their competence for ca-reer concerns, the latter for reelection purposes. The second assumptionis that the tasks for bureaucratic agencies have to be specified ex anteand cannot be contingent on the realization of too many shocks on theenvironment and/or on the public’s preferences. If one accepts thesetwo hypotheses, the nature of our results is quite robust to variations onother less important assumptions.This difference in how incentives play out can then be used at the

constitutional table to design the appropriate allocation of tasks betweenthe two types of policymakers. We have considered three general classesof policies: those concerning a single task, those concerning several inter-related tasks, and those concerning redistributive tasks. Consider firstthe case of a single policy dimension. Tasks requiring special skills andabilities (i.e. skills not shared by the population at large and with highuncertainty about the policymaker’s ability) are better left in the handsof a bureaucrat. The reason is not so much that the bureaucrat may bemore likely to have these skills, but rather that he has stronger incen-tives to show that he indeed has them. On the other hand, imperfectmonitoring weakens the incentives of both policymaker types, and it isnot necessarily an argument against bureaucratic delegation.Next, consider policies with multiple dimensions, i.e. where the pol-

icy in question has several effects that have to be traded off againsteach other, and where there are many options to choose from. Here,the politician generally is preferable for two reasons. First, it may bedifficult to spell out ex-ante a well defined policy goal and while thisex-ante definition is necessary only for the bureaucrat, on the contrarythe politician is automatically inclined to do whatever is in the inter-est of the voters. Second, the incentives of the politician induce him totake into account policy complementarities (he has to look at the overallwelfare effect on voters), and to compensate the losers. A bureaucratinstead has a narrower vision due to the specificity of the task assignedto him. These advantages of the politician become a handicap in thecase of time inconsistent policies, however. Here, a narrowly defined bu-reaucratic mission enhances policy credibility and becomes a reason toprefer a bureaucrat.Finally, consider redistributive policies. Here the nature of the con-

flict matters. If the policy creates a conflict between voters at largeagainst powerful but small special interests, then political delegation ismore risky, because the politician is more likely to be captured by theinterest group compared to the bureaucrat. The reason is that the in-terest group has one more weapon to use, with the politician, namelyto help him be reelected; this weapon is useless instead in the case of

27

the bureaucrat. The conclusion is reversed, however, in the case of re-distributive conflicts between large groups in the population. Here it isgenerally impossible to clearly define the bureaucratic mission in termsof the interest of the voters’ at large; hence, there is always the risk thatthe bureaucrat will arbitrarily favor one group over the other. Moreover,even if this risk could be avoided, the constitution may not have enoughcommitment power to prevent removing the bureaucrat from office toreplace him with a politician that would benefit the majority of voters.Hence, even if desirable, bureaucratic delegation may be unfeasible.Our framework can be extended in several directions. One relates

to the endogenous choice of procedures, especially with regard to theirtrasnparency. Politicians may not always prefer the most transparentway of presenting their acts, an issue explored also in Alesina and Cukier-man (1991) in a different context.22 One of the most often cited examplehas to do with budgetary procedures. Often public budgets are exceed-ingly complicated, and these complicated budget documents are ap-proved in very cumbersome ways.23 This complication is often higherthan necessary, perhaps to make it harder for the public at large to dis-cover all the various favors made to pressure groups in the darker cornersof the budget. If the degree of transparency (say the variance of ε in ourcase of imperfect monitoring above) can be chosen by the policymaker,both the politician and the bureaucrat may prefer opaque procedures.An optimal arrangement here might be to split responsibilities, so thatthe policymaker choosing the procedure is not the same one in chargeof making decisions under that procedure (i.e., he is not the residualclaimant of the rents induced by a less transparent procedure).24

Finally, we have characterized what would happen if the constitu-tion was written optimally, that is behind a veil of ignorance. But thisis not normally the case. For instance, to the extent that existing politi-cians may have an important role at the table in which Constitutionsare written, one may find that actual Constitutions may deviate signifi-cantly from the optimality criteria that we have sketched. For instancereal world politicians may push towards keeping for politicians preciselythose functions that make their life easier (or richer); for instance politi-cians may be keen on retaining exactly those functions that are likelyto generate generous campaign contributions even though, precisely for

22These authors show that an ”ideological” polticians may prefer to have his poli-cies less observable to appear more moderate in a run up to the election.23See the volume edited by Poterba and von Hagen (2000) for an extensive discus-

sion of the role of transparency in the budget process.24Persson, Roland and Tabellini (1987) study the benefits of separation of powers

in a related context.

28

these reasons, these functions should be delegated to bureaucrats. Thus,while we have emphasized a ”normative” analysis of the ConstitutionalTable, one could use our framework to discuss what politicians wouldchoose to allocate to themselves and what not. This would certainlylead to different constitutional trade-offs compared to those discussed inthe previous pages.

Appendix

1. Time inconsistencyThere are two tasks, i = 1, 2, and:

U(y1, y2) = y1 + y2 (36)

For task two y2 = θ+ a2. But y1 depends also on private sector expec-tations, ae2.

y1 = θ + a1 − (a2 − ae2) (37)

Suppose throughout that policy commitments are unavailable, mean-ing that first private expectations are formed, and then effort in bothtasks, a1 and a2, are chosen. In order to stress the difference betweenthe bureaucrat and the politician, suppose now that costs are additive:c = C(a1) + C(a2). The politician allocates effort so as to maximize:

Pr(y1 + y2 ≥W )− C(a1)− C(a2) (38)

taking the voters’ reservation utility,W, and the private sector expec-tations, ae2, as given. In equilibrium, W = 2θ̄ + ae1 + ae2 . Taking thefirst order optimality conditions for the politician and imposing rationalexpectations, yields the following result:

1

rn(θ̄)=Ca(a

P1 ) (39)

aP2 =0

Equilibrium effort on task 1 is determined by the same condition as insection 3, except that the left hand side is divided by 2 because nowtask 1 only contributes 50% to improve the politician’s chances for re-election. But the politician exerts no effort at all on task 2 becauseex-post the benefit for the voters from this policy outcome are exactlyoffset by the negative effect on the performance of task 1. Since votersassign equal weights to both tasks, and effort is costly, the politician

29

ex-post prefers to do nothing. Of course, this is suboptimal from anex-ante perspective: only unexpectedly high a2 hurts the performance oftask 1, and under rational expectations the voters would be better offif the politician could commit to exert high effort also in task 2, andexpectations were formed accordingly. Overall voters’ utility under thepolitician is thus:

UP = 2θ̄ + aP1 (40)

Next, consider the bureaucrat, and suppose that his ability is evalu-ated according to a composite measure of performance, y = δy1 + (1−δ)y2, as in (11) above. Repeating the same steps, and still taking expec-tations as given, we now obtain:

δ=Ca(aB1 ) (41)

(1− 2δ)≤Ca(aB2 ) (42)

Like the politician, and for the same reasons, the bureaucrat too exertsless effort in task 2 than in task 1, because under discretion he perceivesa cost from unexpectedly high effort. In fact, for δ ≥ 1/2, (42) impliesaB2 = 0. But now, the constitution gives a tool to overcome this incentiveproblem: tilting the bureaucratic mission towards task 2, with δ < 1/2,induces the bureaucrat to reduce aB1 and increase aB2 . Since costs areconvex, at least over some range aB2 increases by more than a

B1 is reduced.

Moreover, if expectations are formed after the constitutional stage, this isreflected into expectations, and aB2 = ae2, so that the loss in performancein task 1 is more than offset by the improved performance in task 2.Hence, the voters’ expected utility is:

UB = 2θ̄ + aB1 + aB2 (43)

Unless effort by the politician in task 1 is very high, the voters are likelyto be better off under the bureaucrat. In fact, voters would be evenbetter off if tasks 1 and 2 could be split between two distinct bureaucrats(or between a politician in charge of task 1 and a bureaucrat in chargeof task 2). The bureaucrat in charge of task 2 could be given a missiondefined only on y2 as a basis of performance, and someone else could bein charge of task 1. This would get rid entirely of the time inconsistency,since the bureaucrat in charge of task 2 would now disregard completelythe negative impact of unexpectedly high a2 in the performance of theother task. The proposition in the text follows.

2. LobbyingAs stated in the text, the equilibrium with lobbying must solve the

following optimization problem by choice of a1,a2 and f, subject to non-negativity constraints on the three choice variables, and taking voters’

30

expectations ae1 as given, as before.

Max©Pr(θ ≥ θ̄ + ae1 − a1 −H(f)) + (1 + γ)a2 − C(a1 + a2)− f

ª(44)

The first order conditions for a1, a2 and f evaluated at the pointae1 = a1 imply respectively:

n(θ̄ −H(f))− Ca(a1 + a2) + µ1=0 (45)

1 + γ − Ca(a1 + a2) + µ2=0 (46)

n(θ̄ −H(f))Hf(f)− 1 + µ3=0 (47)

where µi, i = 1, 2 are the lagrange multipliers on the non-negativity con-straints for ai, while µ3 is the lagrange multiplier on the non-negativityconstraint for f.Consider first the case Hf(0) < 1/(1 + γ). Since Hff < 0, here

lobbying is inefficient, and the first order conditions can only be satisfiedif f = a2 = 0 and a1 is at an interior optimum defined by (45) with µ1 = 0in it.Next, consider the case Hf(f

∗) > 1/(1 + γ). This is the oppositeextreme, in which lobbying is very effective. In this case a1 = 0 anda2 and f∗ are at an interior optimum defined by (46) and (47) withµ2 = µ3 = 0 in them.In the intermediate case, in which Hf(0) > 1/(1+γ) but the returns

to campaign contributions fall rapidly, an equilibrium with lobbying doesnot always exis. A special knife edge case is given by the case in whichHf(0) > 1/(1 + γ) and Hf(f

∗) = 1/(1 + γ) = n(θ̄ −Hf(f∗)). Here a1

and a2 can both be positive, and are defined by

1 + γ = Ca(a1 + a2)

and by the condition that the politician is indifferent between this equi-librium and the one with no lobbying.

31

References

[1] Aghion, P. A.Alesina and F. Trebbi (2002) ”Endogenous polticalinstitutions” NBER Working Paper.

[2] Alesina A., O. Blanchard, J Gali, F. Giavazzi and H. Uhlig (2001)Designing Macroeconomic Policy for Europe, CEPR London

[3] Alesina A.and Cukierman, A. (1991) ”The politics of ambiguity.”Quarterly Journal of Economics 105: 829-850.

[4] Alesina A. and R. Gatti (1995) ”Independent Central Banks: Lowinflation at no costs?” American Economic Review, 85:196-200.

[5] Alesina A. and L. Summers (1993) ”Central bank independence andmacroeconomic performance: Some comparative evidence.” Journalof Money, Credit and Banking 25: 151-62.

[6] Barro R. and D. Gordon (1983) ”Rules, discretion and reputationin a model of monetary policy.” Journal of Monetary Economics12: 101-121.

[7] Besley T. and S. Coate (2003), "Elected versus Appointed Regu-lators: Theory and Evidence", Journal of the European EconomicAssociation, forthcoming.

[8] Besley T. and M. Gathak (2003)” Competition and Incentives withmotivated agents” unpublished.

[9] Blinder A. (1998) Central Bank in Theory and Practice. MIT Press.[10] Blinder A. (1997) Is government too political? Foreign Affairs[11] Business Council of Australia (1999) Avoiding boom/bust: macro

econbomic reforms for a globalised economy Discussion Paper, Oc-tober

[12] Cukierman A. (1992) Central Bank Strategy Credibility and Inde-pendence MIT Press Cambridge Mass.

[13] Dewatripont, M., Jewitt, I., and Tirole, J. (1999a). ”The economicsof career concerns, part I: Comparing information structures.” Re-view of Economic Studies 66, 183—198.

[14] Dewatripont, M, Jewitt, I., and Tirole, J. (1999b). ”The economicsof career concerns, part II: Application to missions and accountabil-ity of government agencies, Review of Economic Studies 66, 199—217.

[15] Dewatripont M.and J. Tirole (1999) ’Advocates” Journal of PolticalEconomy, 107: 2-39

[16] Epstein D. and S, O Halloran (1999) Delegating Powers CambridgeUniversity Press, Camnbridge UK

[17] Ferejohn, J. 1986. ”Incumbent performance and electoral control.”Public Choice 50: 5-26.

[18] Fiorina M. (1977) Congress: the Keystone of American DemocracyYale University Press, New Haven CT

32

[19] Grilli, V., Masciandaro, D., and Tabellini, G. (1991).”Political andmonetary institutions and public financial policies in the industrialcountries.” Economic Policy 13: 342—392.

[20] Grossman E. and G. Helpman (2001) Special Interest Poltics MITpress Cambridge Mass

[21] Holmstrom B.(1999) ”Managerial Incentive Problems: A dynamicPrespective” Review of Economic Studies, 66.

[22] Kydland, F., and Prescott, E. 1977. ”Rules rather than discretion:the inconsistency of optimal plans.” Journal of Political Economy85: 473-490.

[23] Lowi T.(1969) The end of liberalism: the second Republic of theUnited States Norton, New York

[24] Mc Cubbins M. R. Noll and B. Weingast (1987) ”AdministrativeProcedures as Instruments of Poltical Control” Journal of Law Eco-nomic and Organization” 3:243-77

[25] Mc Cubbins M. R. Noll and B. Weingast (1989) ”Structure andProcess Poltics and Policy: Administrative Arrangements and thePoltical Control of Agencies” Virginia Law Review, 75:431-82.

[26] Maskin E. and J. Tirole (2001) ”The judges and the politicians”unpublished.

[27] Persson T, and G. Tabellini (1990)Macroeconomic Policy Credibiltyand Politics Chur: Arwood Academic Publishers

[28] Persson T, G. Roland and G. Tabellini ( 1997) ”Separation of pow-ers and political accountability.” Quarterly Journal of Economics112: 1163—1202.

[29] Persson T. and G. Tabellini (1993) "Designing institutions for mon-etary stability" Carnegie-Rochester Conference Series on PublicPolicy 39: 53-89.

[30] Persson T. and G. Tabellini (2000) Political Economics. ExplainingEconomic Policy, MIT Press.

[31] Putnam R. (1988) ”Diplomacy and Domestic Politics: The logc oftwo level games” International Organizations 427-81

[32] Poterba, J., and von Hagen, J (eds.). (1999). Fiscal Rules and FiscalPerformance.Chicago: University of Chicago Press.

[33] Rogoff, K. (1985). ”The optimal degree of commitment to an in-termediate monetary target.” Quarterly Journal of Economics 100:1169-1190.

[34] Rogoff K. (1990) ”Equilibrium Political Budget Cycles” AmericanEconomic Review,80:21-36

[35] RogoffK and A. Sibert (1988) ”Elections andMacroeconomic PolicyCycles” Review of Economic Studies 55: 1-16.

[36] Seabright, P. 1996. ”Accountability and decentralization in govern-

33

ment: An incomplete contracts model.” European Economic Review40: 61-89.

[37] Stigler G. (1971) ”The Theory of Economic Regulation” Bell Jour-nal of Economic and Management Science 2: 3-21.

[38] Vreeland J. 2003 The IMF and Economic Development, CambridgeUniversity Press

[39] Walsh, C. 1995. ”Optimal contracts for central bankers.” AmericanEconomic Review 85: 150-167.

[40] Wilson J.Q. (1989) Bureaucracy Basic Books, New York.

34